00:46:13 [QUIT] ink[sleep] quit: Ping timeout for ink[sleep][ppp-207-214-212-83.sntc01.pacbell.net]
02:32:32 Rhamphoryncus joined #tunes
02:44:41 jemarch joined #tunes
02:44:48 <jemarch> COME TO TALK WITH MEGAHAL (A CONVERSATION SIMULATOR) IN #MEGAHAL
02:44:50 jemarch left #tunes
02:48:48 Kyle joined #tunes
02:52:23 <Kyle> Rham, are you up?
02:52:56 [QUIT] pratap quit: ircII EPIC4-2000 -- Accept no limitations
02:54:11 <Rhamphoryncus> yup
02:56:16 <Kyle> I read you like compilers, and optimization.
02:56:25 <Rhamphoryncus> yeah
02:56:33 Rhamphoryncus is a junkie for perfection )
02:57:14 <Kyle> I do not have the stomach for optimization.
02:58:07 <Rhamphoryncus> you mean making them?  and do you mean sourcecode, or in compilers?
02:58:51 <Kyle> Both.  That is why I like Tunes, I hope it will hide my littel secret from other developers.
02:58:58 <Rhamphoryncus> hehe
02:59:07 <Rhamphoryncus> I hate.... almost everything about programming
02:59:17 <Rhamphoryncus> I just like thinking about stuff
02:59:53 <Kyle> Same here.  I think you have describe about everyone here.
03:00:03 <Rhamphoryncus> hehehe
03:01:08 <Kyle> Have you found a way to characterize optimization (for compilers)?
03:01:17 <Rhamphoryncus> I'm not sure what you mean
03:02:14 <Kyle> Can all optimizations be shown as a graph rewrite, or are there cases where that is not an efficient semanic use.
03:02:27 Rhamphoryncus blinks
03:02:44 <Rhamphoryncus> sorry, but I still don't understand (I sorta do, but not completely)
03:03:09 <Kyle> I would suppose that the compiler starts with a parse tree.
03:03:15 <Rhamphoryncus> yup
03:03:34 <Kyle> Then You can use rules on the parse tree to eventually reduce it to sets of instructions.
03:03:42 <Rhamphoryncus> yup
03:04:10 <Kyle> This is a specialization of graph rewrite, changing the "structure" without changing the meaning
03:04:16 <Rhamphoryncus> ahhh
03:04:46 <Rhamphoryncus> well then to answer your question, I have alot of ideas on how to do it, but I have no exact method that I know will work
03:05:20 <Kyle> I have a compiler that needs optimization.  So I wonder about the best way of adding optimization schemes to it.
03:05:34 <Rhamphoryncus> hmmm
03:05:54 <Rhamphoryncus> well, *MY* choice would be to just make it into a parser for my compiler, but I'm somewhat biased ;)
03:06:56 <Kyle> What would you do if you "discovered" a new optimization.
03:07:08 <Kyle> albeit small
03:07:08 <Rhamphoryncus> add it to my compiler
03:07:19 <Kyle> as a sub compiler?
03:07:29 <Kyle> preprocessor?
03:07:58 <Rhamphoryncus> well it would have to fit into the existing structure, otherwise it would be of no use
03:08:35 <Kyle> you lost me "existing structure"?  of the compiler, or the parse tree.
03:08:59 <Rhamphoryncus> and it's not really possible to have a preprocessor per-say.  my compiler starts at the parse tree, so there's nothing to preprocess before that
03:09:36 <Rhamphoryncus> well I have a rather large and complex system that the compiler uses, and if it can't be fit into that system it can't be part of my compiler
03:09:39 <Kyle> You are proposing many little parsers that "jump in" to help compile to machine code.
03:09:56 <Rhamphoryncus> no... 
03:11:16 <Rhamphoryncus> btw, do you know how to copy and paste in vim?  I can never remember
03:11:25 <Kyle> no
03:14:54 <Kyle> do you have docs that I can read so I don't ask too many dumb questions?
03:14:54 <Rhamphoryncus> some, but they're really crappy right now
03:14:54 <Rhamphoryncus> http://rhamphoryncus.dyndns.org/~rhamph/Foundation/F1/
03:14:54 <Kyle> thanks.
03:14:54 <Rhamphoryncus> careful of the file dates
03:14:54 <Rhamphoryncus> most of the stuff in there is old
03:14:54 <Rhamphoryncus> compiler description.html, description.html, and interface.h are the up to date stuff
03:15:01 <Kyle> have you decided on a parse tree?
03:15:24 <Rhamphoryncus> you mean what form it'll take?
03:15:36 <Kyle> yes, a definition, even if vague
03:15:49 <Rhamphoryncus> to some degree, but not completely
03:16:06 <Rhamphoryncus> interface.h defines the functions the parser will use to create it though
03:17:24 Rhamphoryncus decides it's easier (and much quicker) to just retype the half dozen lines
03:20:20 <Rhamphoryncus> pseudoform.txt is what I'm working on at this moment.  trying to make it follow the compiler definition.html
03:21:15 <Kyle> I find it easier to read what a person is doing by seeing thier datastructures than thier functions.
03:21:29 <Rhamphoryncus> true
03:22:11 <Rhamphoryncus> most of the docs were written primarly for my benefit, and therefor focus on things I was working on at the time
03:22:43 <Rhamphoryncus> description.html is written for other people's benefits.  unfortunetely, I was having alot of trouble getting work done on it
03:26:14 <Rhamphoryncus> one thing I might do at a later point is hook up something that uses the same interface that outputs a diagram instead of compiling
03:27:12 <Kyle> that would be good.  I have an auto html class describer for my system
03:27:41 <Rhamphoryncus> well it wouldn't be good for much real work
03:28:37 <Kyle> make sure to script it in a language where it will not take much time.
03:28:50 Rhamphoryncus mumbles
03:29:13 <Kyle> maybe C has libraries that make the job easier.
03:29:18 <Rhamphoryncus> you know, I still have to come up with the parser
03:29:18 <Kyle> i don't know
03:29:25 <Kyle> )
03:29:53 <Rhamphoryncus> I want to make the thing self hosting, but it's not actually capable of being self hosted
03:30:08 <Rhamphoryncus> it needs a language with a written syntax first
03:30:20 <Kyle> "self hosting"?
03:30:32 <Rhamphoryncus> I want to write the real compiler in the language itself
03:30:50 <Kyle> We all want to do that.
03:31:10 <Rhamphoryncus> but it's hard to do that with a language that just provides function calls :)
03:31:20 <Kyle> Are you choosing a syntax, or making one up.
03:31:29 <Rhamphoryncus> making one up
03:31:33 <Rhamphoryncus> might as well
03:32:24 <Kyle> How is it hard.  I would imagine the simplicity of the paradigm would make it easy.
03:32:58 <Rhamphoryncus> well you can't exactly trigger an interrupt by hand.  you write a program that does it
03:33:33 <Kyle> i think you lost me.
03:33:44 <Rhamphoryncus> heh
03:33:53 <Rhamphoryncus> not surprising.  this thing confuses everybody :)
03:34:24 <Rhamphoryncus> my language does not provide a parser.  my language is *just* the set of functions given in interface.h
03:35:32 <Kyle> MMmmm
03:36:43 <Kyle> How do you create a function that adds two numbers?
03:37:07 <Rhamphoryncus> well first you create the numbers... btw, I don't have any integers either.
03:37:23 <Kyle> that is consistent with tunes
03:39:26 <Rhamphoryncus> then you use a couple iterators to go over the each vector of bits and add them together, the same way a CPU would.  (well, most CPUs would do them all simultaneously)
03:40:49 <Kyle> Mmmm
03:42:25 <Rhamphoryncus> actually, semantics would give you a better idea of how things work.  it's just a bit out of date though
03:42:36 <Kyle> yes
03:42:55 <Kyle> Well, I have to go to work now.   
03:42:59 <Rhamphoryncus> ok
03:43:01 <Rhamphoryncus> ttyl
03:43:01 <Rhamphoryncus> )
03:43:05 <Kyle> Thanks for the conversation.
03:43:09 <Rhamphoryncus> np
03:43:13 [QUIT] Kyle quit: Leaving
05:43:31 karltk joined #tunes
05:48:11 Fare joined #tunes
07:53:49 hcf joined #tunes
07:55:27 <hcf> abi: perligata is Lingua::Romana::Perligata, Perl for the XXI-imus Century, write Perl in Latin, largely non-positional syntax, at http://www.csse.monash.edu.au/~damian/papers/HTML/Perligata.html
07:56:37 [QUIT] Fare quit: Ping timeout for Fare[ppp66-net1-idf2-bas1.isdnet.net]
07:57:06 Fare joined #tunes
09:08:17 <hcf> abi: cuecat.scm is at ftp://ftp.meme.com/pub/software/cuecat/
09:43:43 [QUIT] JALH quit: brb
09:59:54 eihrul joined #tunes
09:59:59 eihrul hmms.
10:04:25 <eihrul> Kyle (logged): a lot of optimizations can be classified as either data-flow optimizations and most of the rest are simple graph rewrite
10:04:56 <eihrul> whereas one collects data about nodes and makes decisions about what to do with them based on the collected data, the rest are just simple reductions :)
10:16:54 [QUIT] eihrul quit: Leaving
10:24:40 [QUIT] hcf quit: Leaving
10:39:06 [QUIT] MysticOne quit: See ya!
10:39:27 [QUIT] HolyK quit: Ping timeout for HolyK[194.126.30.178]
11:06:07 [QUIT] lar[sleep] quit: Ping timeout for lar[sleep][adsl-63-204-133-55.dsl.snfc21.pacbell.net]
11:06:14 lar[sleep] joined #tunes
11:08:35 Rhamph-Sleeping joined #tunes
11:21:54 eihrul joined #tunes
12:08:19 [QUIT] JALH quit: Ping timeout for JALH[user38-35.jakinternet.co.uk]
12:21:07 moebious joined #tunes
12:24:39 <moebious> how goes?
12:26:21 Fare joined #tunes
12:43:26 <moebious> what do  you need to get a system such as TUNES running on your computer?  IE, what will be the assembly code that start the whole system the first time its loaded? 
12:44:03 <moebious> An interpreter and an entry point into the codebase?
12:44:21 <moebious> but you would also need drivers, for the filesystem, ect
12:52:55 [NICK] _ruiner_ changed nick to: cog
13:38:24 <Fare> hum
13:38:46 <Fare> why should it be different than other systems, in this respect?
13:39:06 <Fare> you have a multi-step boot loader, that loads an image into memory, etc.
13:39:22 <Fare> actually, this isn't the most interesting part of the system
13:39:24 [QUIT] JALH quit: g'night *
13:39:44 <Fare> one interesting way to develop it would be to metaprogram it in a declarative way
13:40:18 <Fare> but since you typically don't want to touch this part often, it'd be easier to just steal another OS'es boot loader (e.g. GRUB or LILO)
13:40:43 <eihrul> either that or just have have objects in your language that expose enough low level details to make dealing with those issues as easy as any other objects :P
13:42:29 <eihrul> i.e. Squeak has Slang, for that purpose
13:58:08 <moebious> eihrul, thats more what I was thinking was intended.  I had this idea with "contexts" (my "objects") that I think could do this, I think.
14:00:38 <eihrul> normal objects can do this just fine :)
14:00:57 <moebious> basicly, a section of the os is the machine definition, which among other things, has some basic (essential) translations of whatever language is in use to machine code, as well as some example code the machine could work ...
14:00:58 <moebious> with.  My cheat was to have an example interpeter here, but have that loaded at boot time, and enter at the top level somewhere.
14:01:32 <moebious> In time, the system could replace it with its own interpeter, perhaps, since the the interpeter would be concurently defined in the HLL :)
14:02:08 <eihrul> that's called bootstrapping, nothing new
14:02:20 <moebious> eh, yea
14:03:25 <moebious> thats not exacly what most os's bootstrap however.
14:04:06 <eihrul> eh? how is that not just bootstrapping?
14:04:34 <moebious> i didn't say it wasn't.  define bootstrapping
14:05:37 <eihrul> incremental replacement of external primitives with internal ones :)
14:06:04 <moebious> trust me, i'm speaking out of my ah,.. out of a wide base of knowledge ;)
14:06:35 <moebious> define primitives
14:07:15 <eihrul> primitive being something not expressed (and possibly not expressable) internally
14:07:40 <moebious> ok, makes sence
14:08:06 <moebious> i always thought it just loaded a kernel and started exicution somewhere.
14:08:14 <moebious> sorry about the spelling.
14:08:15 <eihrul> you can do that too
14:08:24 <eihrul> but the boot loader is effectively your boot strap mechanism, in that case
14:10:04 <moebious> i'm not saying I am proposing anything new.  I haven't read the docs, this is just what I came up with over a cup of coffee, based on simpling knowing a little bit of what you/we are trying to do.
14:10:27 <eihrul> these problems are already understood though :)
14:10:33 <eihrul> they're not really issues at this point
14:10:38 <moebious> hehe
14:15:43 <moebious> I was thinking of the basic, eh, "object" being a context.  A context acts like a barrior, in that it imposes restrictions on everything which spawns from it (not sure how this would work), thats the important thing.  Every ...
14:15:43 <moebious> context is produced as a "child" of sorts under the current one.
14:16:32 <eihrul> in other words, inheritance
14:16:55 <moebious> i dont think its the same
14:17:25 [QUIT] cog quit: Killed (NickServ (This nick is reserved by another user))
14:19:46 <eihrul> why isn't it?
14:21:18 <moebious> its possable i just thought of a different way for ME to think about inheritance.  I'm trying to find a clear conter example.
14:21:31 <moebious> you mean inheritance in the sence of classes and objects?
14:21:46 <eihrul> inheritance in the sense of objects
14:21:56 <eihrul> classes are just a conceptual tool for thinking about inheritance :)
14:22:03 <eihrul> but not necessary to express it
14:24:06 <moebious> yes, well
14:24:33 <moebious> i cant think of a counter example, i think they are the same.
14:24:47 moebious slumps.
14:25:30 <moebious> well, its back to the cafe ...
14:28:31 [NICK] kc-teriyaki changed nick to: kc5tja
14:36:03 <moebious> i just tended to think of an object as something which had an address, was an instance of some class, but, I am wrong
14:36:33 <moebious> damn that c++
14:36:38 <moebious> )
15:26:05 Brianna joined #tunes
15:26:54 [QUIT] Fare quit: Ping timeout for Fare[195.154.50.66]
15:29:47 Fare joined #tunes
16:23:36 [NICK] lar1 changed nick to: lar[game]
17:28:45 eihrul joined #tunes
17:39:36 [QUIT] MysticOne quit: Ping timeout for MysticOne[porpoise18.panama.gulf.net]
17:40:49 [QUIT] Melinda quit: Ping timeout for Melinda[porpoise18.panama.gulf.net]
18:12:30 Kyle joined #tunes
18:23:33 <Kyle> is eihrul here?
18:23:40 <eihrul> yeah
18:24:56 <Kyle> data-flow optimizations... do you have an example?
18:26:15 <eihrul> well, type inference is a good example :)
18:26:51 <eihrul> other things like constant propagation are also good examples :)
18:28:33 <Kyle> I will think about that....  thanks for now.
18:35:48 <Kyle> is there a finite, known, set of data-floaw optimizations?  
18:36:05 <Kyle> if not, is there a way to describe large classes of them?
18:36:37 <Kyle> It would be unfortunate to program these optimizations, programs are hard to parse and understand.
18:38:22 water joined #tunes
18:38:36 <Kyle> hi water
18:38:41 <water> hey
18:39:41 <water> re all
18:39:47 <water> hm, another new person
18:39:50 [QUIT] kc5tja quit: THX QSO ES 73 DE KC5TJA/6 CL ES QRT AR SK
18:39:56 <water> is anything up atm?
18:40:07 <eihrul> Kyle: sure, there's frameworks for doing all sorts of data-flow optimizations
18:40:16 water waits for his email
18:40:26 <eihrul> Kyle: because collecting the data is largely similar, however restructuring done because of data-flow optimization is not necessarily similar
18:40:33 water checks logs ;)
18:40:44 <eihrul> although there's lots of problems where splitting and reorganizing is done uniformly to improve data-flow information
18:42:46 <Kyle> I am not really interested in the data-flow information, just the actions taken based on the info.
18:42:47 water reels at the volume of email in the last 34 hours
18:42:55 [QUIT] Fare quit: Leaving
18:45:16 <water> heh
18:45:34 <water> um... my resume just got forwarded by that friend of mine
18:46:21 <eihrul> Kyle: those are widely different
18:46:49 <Kyle> "those"?
18:47:42 <eihrul> those => the actions taken based upon the data-flow information
18:47:47 <Kyle> sorry
18:48:02 <eihrul> there's no one silver bullet to make a perfectly optimized programs
18:48:09 <Kyle> too bad
18:48:32 <eihrul> because optimization is an art, its mainly heuristic :)
18:48:34 <Kyle> I will leave it to the experts.  Too much work in that field. ;)
18:48:44 <water> yeah there is, "don't run the program" :)
18:49:12 <eihrul> water: that makes the program infinitely slow!
18:49:22 <eihrul> )
18:49:30 <water> no, wait until the hw improves :)
18:51:07 water relishes the hope of these darpa research people getting him out of the navy
18:53:36 [NICK] lar1 changed nick to: lar[dinner]
18:54:53 [QUIT] Aardappel quit: Ping timeout for Aardappel[ipd54b5f9c.free.wxs.nl]
18:55:41 <water> damn it, i must make a phone call
18:55:43 <water> brb
18:56:31 [QUIT] Kyle quit: Leaving
19:05:05 <water> wow my life could get pretty busy soon
19:05:28 <eihrul> Kyle: there's good books on the subject :)
19:06:59 <water> brb again
19:10:02 <water> geez
19:10:12 <water> the switch is apparently too busy
19:20:36 [NICK] lar[dinner] changed nick to: lar1
19:27:47 [QUIT] abi quit: irc.linux.com varley.openprojects.net
19:27:47 [QUIT] clog quit: irc.linux.com varley.openprojects.net
19:27:47 [QUIT] ult quit: irc.linux.com varley.openprojects.net
19:37:30 [QUIT] karltk quit: Ping timeout for karltk[janus.prosalg.no]
19:39:12 karltk joined #tunes
20:02:12 ult joined #tunes
20:02:45 [NICK] lar1 changed nick to: lar[away]
20:06:23 Kyle joined #tunes
20:06:35 <water> re kyle
20:06:42 <Kyle> 'lo
20:06:49 moebious needs to spend a good year or so reading before attempting further discussion
20:06:56 <water> lol
20:07:24 <water> mobius: yep, that is tunes' curse :)
20:07:37 <eihrul> maybe a few months
20:07:40 <moebious> That would be for the purposes of "not talking from ass" :)
20:07:41 <eihrul> not necessarily a year
20:07:54 <moebious> optimism...
20:08:06 <water> no, that's experience
20:09:50 <moebious> thats my curse, i like trying to do everything from scratch.  like, i was messing with the four color theorm for over a year, still sorta am, silly as it is.  refused to ever read about the theorm, ie, what mistakes others ...
20:09:51 <moebious> made.
20:10:32 [QUIT] Kyle quit: Leaving
20:12:08 <moebious> What do you people mean by objects?
20:12:40 <water> something accessed by a symbol, perhaps
20:23:09 <moebious> I think my contexts had some conceptual difference from objects.  Generally, people inherit from an object to expand on its capabilities.  However, contexts work on the concept of also adding restrictions to all further ...
20:23:10 <moebious> "children" 
20:23:33 <eihrul> well, if you consider expanding on capabilities to be refining them
20:23:37 <eihrul> then its same diff :)
20:23:38 water scratches his hed
20:23:46 <water> yeah basically
20:26:04 <moebious> well thats what i realised earlier.  really, i was just looking at a sort of paridigm, and a choice of charictoristics for the "head object", or whatever it is, that everything inherits from.
20:26:45 <water> the root object idea does have some drawbacks
20:27:19 <water> mostly that you wind up with features that are not factored well
20:27:48 <eihrul> well, just because its the root object doesn't mean it can't inherit from others :)
20:27:49 <moebious> i thought, in practicality, that it would only be the root object "on paper" so to speak, just so there is a cool looking definition of everything in the language
20:28:05 <water> ok
20:28:19 <eihrul> water: you just have longer cycles :)
20:28:53 <water> heh sure
20:29:40 <water> i just meant that some of Object's features in most languages have to be put there for practical purposes, but really don't belong there
20:29:53 <water> slate doesn't have that concern
20:41:55 [QUIT] Melinda quit: Read error to Melinda[porpoise21.panama.gulf.net]: Connection reset by peer
20:43:26 <water> wow my life is getting busy
20:43:38 <eihrul> with? VC? contacts?
20:43:59 <water> job interviews also
20:44:07 <water> and i'm investing in a company
20:44:28 <water> not publically traded, and i can't talk about it
20:44:55 <eihrul> so this could possibly be the leak through with the water escapes the ship? :)
20:45:05 <water> we'll see
20:45:22 <eihrul> s/with/which
20:46:53 <water> i haven't even started up squeak yet!
20:47:03 water smacks his forehead
20:47:19 <water> otoh i did sneak in a lot of sleep at work ;)
20:47:24 <eihrul> good thing i made a rhyme
20:47:27 <water> so i'll be up late
20:47:42 <water> what rhyme?
20:47:49 <eihrul> leak and squeak
20:48:01 <eihrul> and possibly sneak
20:48:07 <water> heh
20:49:54 <water> why is that a good thing?
20:51:32 <eihrul> you never would have remembered
20:51:43 <water> ??
20:51:59 <eihrul> never mind, too late for that
20:52:23 <water> uh i know why the leak name was useful
20:54:15 eihrul goes off to sleep before water tries to read more into a bad rhyme.
20:54:28 <water> yeah that's a good idea
20:54:45 water is about to get into some heavy functional pattern hacking
20:55:11 [QUIT] eihrul quit: Leaving
20:55:13 hcf joined #tunes
20:55:20 <water> hey hcf
20:59:13 abi joined #tunes
20:59:13 clog joined #tunes
21:13:30 [QUIT] hcf quit: Leaving
21:29:24 [QUIT] MysticOne quit: Ping timeout for MysticOne[porpoise21.panama.gulf.net]
21:42:38 <water> hi
21:43:15 <Rhamphoryncus> hiya
21:43:35 water read the logs
21:59:20 water makes some minor updates to his homepage
22:12:35 <moebious> the other day, i got kicked from the #freebsd on efnet for trying to make conversation.
22:12:47 <moebious> msg was "do we care?"
22:13:05 <water> about what?
22:13:38 <moebious> my comment was something like "I have a 386 I want to put BSD on.  Why?  I dont know."
22:14:10 <moebious> i like bsd, but they dont do a good job at PR for it there.
22:14:15 [NICK] lar[away] changed nick to: lar1
22:14:32 <water> oh well
22:16:00 <moebious> so thats why we need to get TUNES working, pronto =]
22:16:15 <water> hm
22:16:58 <water> well, i'm still working on this code in squeak, which barely supports self-style objects
22:17:07 <water> trying to make it work for slate
22:17:13 <moebious> "TUNES, make me a new application, one that does spreadsheet."  "Yes user Derek.  I will start now."  right??? :P
22:17:39 <water> *sigh*
22:17:45 <moebious> hehe
22:18:02 <water> squeak has a spreadsheet class
22:18:07 <moebious> ah
22:18:09 <water> with a gui
22:18:23 <water> its not excel or gnumeric
22:18:35 <water> but you can extend it very easily
22:18:50 <moebious> actually, thats kinda amazing, since I didn't hear about squeak untill now.
22:19:20 <water> common lisp has a lot of good packages, too, but is less approachable
22:20:56 <moebious> many decent gui packages?  I haven't delved that deep.
22:21:53 <water> for which?
22:22:13 <moebious> lisp i ment.
22:22:18 <water> oh
22:22:20 <moebious> common lisp.
22:22:32 <water> afaik lisp doesn't have a self-hosted gui like smalltalk does
22:24:22 moebious thinks about lisp to c bindings.
22:24:44 <water> squeak already has a binding called slanf
22:24:49 <water> oops... slang
22:25:21 <moebious> neat.  need to learn squeak i guess.  talk to you later
22:25:26 <water> ok
22:27:01 moebious left #tunes
22:36:43 [QUIT] alterego quit: [BX] Get your free warez from ftp://127.0.0.1!
00:26:07 [QUIT] alterego quit: BitchX-1.0c16 by panasync